Published on Web 11/05/2003
Crystal Structure and Hydrogen Bonding System in Cellulose Ir from Synchrotron X-ray and Neutron Fiber Diffraction Yoshiharu Nishiyama,† Junji Sugiyama,‡ Henri Chanzy,§ and Paul Langan*,|
Downloaded via IOWA STATE UNIV on January 16, 2019 at 03:24:44 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
Contribution from the Department of Biomaterial Sciences, Graduate School of Agriculture and Life Science, the UniVersity of Tokyo, Yayoi 1-1-1, Bunkyo-ku, Tokyo 113, Japan, Wood Research Institute, Kyoto UniVersity, Uji, Kyoto 611-0011, Japan, Centre de Recherches sur les Macromole´ cules Ve´ ge´ tales, CNRS, affiliated with the Joseph Fourier UniVersity of Grenoble, BP 53, 38041 Grenoble Cedex 9, France, and Bioscience DiVision, Los Alamos National Laboratory, Los Alamos, New Mexico 87545 Received July 3, 2003; E-mail:
[email protected] Abstract: The crystal and molecular structure, together with the hydrogen-bonding system in cellulose IR, has been determined using atomic-resolution synchrotron and neutron diffraction data recorded from oriented fibrous samples prepared by aligning cellulose microcrystals from the cell wall of the freshwater alga Glaucocystis nostochinearum. The X-ray data were used to determine the C and O atom positions. The resulting structure is a one-chain triclinic unit cell with all glucosyl linkages and hydroxymethyl groups (tg) identical. However, adjacent sugar rings alternate in conformation giving the chain a cellobiosyl repeat. The chains organize in sheets packed in a “parallel-up” fashion. The positions of hydrogen atoms involved in hydrogen-bonding were determined from a Fourier-difference analysis using neutron diffraction data collected from hydrogenated and deuterated samples. The differences between the structure and hydrogenbonding reported here for cellulose IR and previously for cellulose Iβ provide potential explanations for the solid-state conversion of IR f Iβ and for the occurrence of two crystal phases in naturally occurring cellulose.
Introduction
Although the crystal structure of cellulose has been investigated for almost a century,1-5 certain details remain to be deciphered. The molecule in cellulose is a linear polymer of β(1f4) linked D glucosyl residues. In native cellulose (cellulose I), these molecules are synthesized in a continuous fashion by terminal complexes (TC) containing a number of cellulose synthases assembled in biological spinnerets at the cell membrane. Each TC spins a crystalline microfibril consisting of parallel, hydrogen-bonded molecules. In some plants, such as ramie or flax, the microfibrils are generated and assembled in the direction of the plant fiber. Such fibers yield X-ray diffraction patterns with particularly high orientation and resolution. Fiber diffraction studies of these samples have provided crystal structures for cellulose I.6,7 However, these structures have been unable to account for all of the diffraction features recorded from samples of oriented cellulose I from algae.8 † Department of Biomaterial Sciences, Graduate School of Agriculture and Life Science, the University of Tokyo. ‡ Wood Research Institute, Kyoto University. § Centre de Recherches sur les Macromole ´ cules Ve´ge´tales, CNRS, affiliated with the Joseph Fourier University of Grenoble. | Bioscience Division, Los Alamos National Laboratory.
(1) (2) (3) (4) (5) (6) (7)
Nishikawa, S; Ono, S. Proc. Tokyo Math. Phys. Soc. 1913, 7, 131. Meyer, K. H.; Mark, H. Ber. Dtsch. Chem. Ges. 1928, 61, 593. Meyer, K. H.; Misch, L. HelV. Chim. Acta 1937, 20, 232. Gardner, K. H.; Blackwell, J. Biopolymers 1974, 13, 1975. Sarko, A.; Muggli, R. Macromolecules 1974, 7, 486. French, A. D. Carbohydr. Res. 1978, 61, 67. Woodcock, C.; Sarko, A. Macromolecules 1980, 13, 1183.
14300
9
J. AM. CHEM. SOC. 2003, 125, 14300-14306
It has subsequently been discovered that the ultrastructure of cellulose possesses unexpected complexity in the form of two crystal phases, namely IR and Iβ.9 IR and Iβ can be found not only within the same cellulose sample,10 but also along a given microfibril.11 The relative amounts of IR and Iβ have been found to vary between samples from different origins. Whereas IR rich specimens have been found in the cell wall of some algae and in bacterial cellulose, Iβ rich specimens have been found in cotton, wood and ramie fibers.12,13 The presence of dimorphism, to a large extent, accounts for the initial difficulties in interpreting X-ray fiber and electron diffraction patterns from algal samples. Unraveling the properties of cellulose I therefore requires the determination of not one, but two crystal structures. However, separating the X-ray data collected from mixed-phase samples into IR and Iβ subsets is hazardous. Fiber samples typically correspond to microcrystallites preferentially aligned along the fiber axis, but with random orientation around this axis. One of the consequences of this cylindrical averaging is that reflection spots tend to overlap in fiber diffractograms, making it almost impossible to reliably separate reflections from different phases. (8) Honjo, G.; Watanabe, M. Nature 1958, 181, 326. (9) Atalla, R. H.; VanderHart, D. L. Science 1984, 223, 283. (10) Sugiyama, J.; Okano, T.; Yamamoto, H.; Horii, F. Macromolecules 1990, 23, 3196. (11) Sugiyama, J.; Vuong, R.; Chanzy, H. Macromolecules 1991, 24, 4168. (12) Horii, F.; Hirai, A.; Kitamaru, R. Macromolecules 1987, 20, 2117. (These authors refer to Ia and Ib to distinguish respectively the ramie and cotton family from that of Valonia and bacterial cellulose). (13) Sugiyama, J.; Persson, J.; Chanzy, H. Macromolecules 1991, 24, 2461. 10.1021/ja037055w CCC: $25.00 © 2003 American Chemical Society
Crystal Structure and H-Bonding System in Cellulose IR
Progress toward resolving this problem has been made in electron diffraction studies of individual cellulose microcrystals extracted from the alga Microdictyon tenuius.11 Diffractograms revealed the presence of two distinct crystal phases, a major component with a one-chain triclinic unit cell and a minor component with a monoclinic two-chain unit cell. The major and minor components were shown to correspond to the IR and Iβ phases previously identified in 13C NMR9 and FT-IR13 studies, respectively. It has also been shown that IR is metastable and can be converted into Iβ by annealing.14 Additional progress has been made with the discovery that tunicin (cellulose from the small sea animal Halocynthia roretzi) consists of nearly pure (∼90%) Iβ phase,15,16 whereas the cellulosic wall of the freshwater alga Glaucocystis consists of nearly pure (∼90%) IR phase.17 Both tunicin and Glaucocystis cellulose consist of highly crystalline microfibrils, but these microfibrils are not sufficiently oriented, as in ramie or flax, for fiber diffraction studies: they have an helical organization in tunicin18,19 and a curved structure in the cell wall of Glaucocystis.20 In view of this difficulty, a method was developed whereby the microfibrils were first hydrolyzed into whiskerlike microcrystals, which were further re-assembled into oriented films amenable to fiber diffraction analyses.21 Samples of both IR and Iβ have been prepared in this manner that diffract synchrotron X-rays to atomic resolution,22 allowing the precise location of carbon and oxygen atoms. A method was also devised for replacing all hydrogen atoms involved in hydrogen bonding by deuterium atoms23 so that their positions could be determined using neutron diffraction.24 The deuterated samples diffract neutrons to resolutions comparable to those obtained with synchrotron X-rays. The determination of the crystal structure and hydrogen-bonding arrangement in cellulose Iβ has already been reported.24 In this work, we describe the crystal structure and hydrogen-bonding arrangement in cellulose IR. Although we report here the first crystallographic structure for cellulose IR, there have been several structures predicted from modeling studies.25-27 The availability, for the first time, of crystallographic coordinates for both IR and Iβ allows a detailed comparison of the structures and hydrogen-bonding arrangements in these two phases. It allows us to discuss possible routes for the solid-state conversion of IR to Iβ. It also contributes toward a scientific basis for addressing the central issue of why these two phases coexist in nature. (14) Horii, F.; Yamamoto, H.; Kitamaru, R.; Tanahashi, M.; Higuchi, T. Macromolecules 1987, 20, 2946. (15) Belton, P. S.; Tanner, S. F.; Cartier, N.; Chanzy, H. Macromolecules 1989, 22, 1615. (16) Larsson, P. T.; Westermark, U.; Iversen, T. Carbohydr. Res. 1995, 278, 339. (17) Imai, T.; Sugiyama, J.; Itoh, T.; Horii, F. J. Struct. Biol. 1999, 127, 248. (18) VanDaele, Y.; Revol, J.-F.; Gaill, F.; Goffinet, G. Biol. Cell 1992, 76, 87. (19) Gubb, D. Tissue & Cell 1975, 7, 19. (20) Willison, J. H. M.; Brown, R. M. Jr. J. Cell. Biol. 1978, 77, 103. (21) Nishiyama, Y.; Kuga, S.; Wada, M.; Okano, T. Macromolecules 1997, 30, 6395. (22) Nishiyama, Y.; Okano, T.; Langan, P.; Chanzy, H. Int. J. Biol. Macromolecules 1993, 26, 279 (23) Nishiyama, Y.; Isogai, A.; Okano, T.; Mu¨ller, M.; Chanzy, H. Macromolecules 1999, 32, 2078. (24) Nishiyama, Y.; Langan, P.; Chanzy, H. J. Am. Chem. Soc. 2002, 124, 9074. (25) Aabloo, A.; French, A. D.; Mikelsaar, R. H.; Pertsin, A. J. Cellulose 1994, 1, 161. (26) Heiner, A. P.; Sugiyama, J.; Telleman, O. Carbohydr. Res. 1995, 273, 207. (27) Vietor, R. J.; Mazeau, K.; Lakin, M.; Perez, S. Biopolymers 2000, 54, 342.
ARTICLES Table 1. Experimental Details X-ray
chemical formula cell setting, space group a(A) b(A) c(A) R(°) β(°) γ(°) V(A3) Z radiation type λ(A) diffractometer independent reflections reflections > 2σ(I) θmax(°) range of h range of k range of l refinement refinement on R[F2 > 2σ(F2)] ωR(F2) ∆Fmax ∆Fmin Frms
crystal data C12H20O10 triclinic, P1 6.717(7) 5.962(6) 10.400(6) 118.08(5) 114.80(5) 80.37(5) 333.3(6) 1 synchrotron X-ray 0.72060
neutron
C12H14D6O10 triclinic, P1 6.717(7) 5.962(6) 10.400(6) 118.08(5) 114.80(5) 80.37(5) 333.3(6) neutron 1.30580
data collection ID2A 255 236 21.10 -6 f 5 -8 f 8 -8 f 8
D19 158 159 16.49 -5 f 5 -8 f 8 -8 f 8
F2 0.1779 0.4296 0.850 -0.768 0.192
F2 0.2049 0.4566 1.552 -1.336 0.414
Experimental Section Glaucocystis nostochinearum, obtained from the IAM culture collection of the University of Tokyo, was cultivated, and its cellulose extracted, purified, and hydrolyzed using previously described techniques.17 After conventional washing by centrifugation followed by redispersion in distilled water, the sample consisted of whiskerlike microcrystals having no more than 20 nm in diameter and several micrometers in length (illustrated in Figure 1a). Each of these microcystals yielded spot electron diffraction diagrams (inserts in Figure 1a) indicating that they corresponded essentially to the triclinic IR phase. In addition, a rotation about the long axis of a given crystal allowed all sections of the reciprocal lattice to be recorded (exemplified by the two inserts in Figure 1a, which are rotated about the crystal axis by 60° with respect to one another). Following a previously described protocol,21,28 the microcrystal suspension was cast as films in which the microcrystals were oriented parallel to one another. These films were used for synchrotron X-ray diffraction experiments, or after deuterium exchange for neutron diffraction experiments. A typical X-ray diffraction diagram is illustrated in Figure 1b. This diagram, together with a number of others, indicates that the samples are not truly symmetric about the fiber axis, but textured. Special strategies developed for textured cellulose Iβ samples were used to collect the X-ray and neutron data from cellulose IR.24 Neutron data were collected from both hydrogenated and deuterated samples (Figure 1c). Reflection positions and intensities were fitted as described previously, 29 with the exception that the orientation of the c* axis, having refined the positions of the a* and b* axes from the equatorial reflection positions, was determined by a grid search using a penalty function Σj(d*o - d*c)2 where d*c is the closest calculated reciprocal reflection spacing from the observed one d*o, and the summation is over all reflections. The experimental parameters are summarized in Table 1. It was found that the exact values of the refined unit cell parameters varied as a function of angle around the fiber axis, probably as a result of internal strain within the constituent films and perhaps also due to slightly ex-centric rotation of the sample during data (28) Nishiyama, Y.; Langan, P.; Chanzy, H. Fiber Diffr. ReV. 2003, 11, 75. (29) Nishiyama, Y.; Langan, P. Fiber Diffr. ReV. 2000, 9, 18. J. AM. CHEM. SOC.
9
VOL. 125, NO. 47, 2003 14301
ARTICLES
Nishiyama et al.
Figure 2. Section through an Omit map calculated using the observed amplitudes and model phases but omitting the hydroxymethyl group oxygen atoms from the phase calculation. The skeletal model represents the cellulose chain. The map is represented at two contour levels in blue and pink. Density (indicated by arrows) can be clearly associated with the hydroxymethyl group in the tg position.
Figure 1. Ia: Electron micrograph of cellulose microcrystals resulting from the acid hydrolysis of Glaucocystis cell wall. When microcrystal areas such as the circled one are probed by electron diffraction, each area yields spot diffraction diagrams (inserts) indicating that the crystals diffract in the IR triclinic system. The two inserts in (a) were obtained by rotating the crystal by 60° about its long axis. Ib: (Top) Synchrotron X-ray diffraction data collected on an online MAR image plate from fibers of Glaucocystis cellulose IR on station ID2A at the ESRF, Grenoble, France. (Bottom) A 3D fit of the Bragg intensities, done using custom-written software that takes into account fiber texture. The images have been remapped into cylindrical reciprocal space with the fiber axis vertical. Ic: (Top) Neutron fiber diffraction patterns collected from two fibers of Glaucocystis cellulose IR, one hydrogenated (left-hand quadrant) and the other deuterated (righthand quadrant). The bottom quadrants show 3D fits of the Bragg intensities, done using custom-written software that takes into account fiber texture. The images have been remapped into cylindrical reciprocal space with the fiber axis vertical. 14302 J. AM. CHEM. SOC.
9
VOL. 125, NO. 47, 2003
collection. The magnitude of this variation is a fraction of a percent for all unit cell parameters and the unit cell parameters reported in Table 1, are the averages. Structure Refinement. X-ray structure refinement was carried out using previously described strategies for applying SHELX-9730 to highresolution fiber diffraction data.31 Atomic starting positions were taken from the average chain conformation in the crystal structure of cellulose Iβ 24, and the chains were arranged in a “parallel-up” fashion.32 An initial refinement was carried out with the hydroxymethyl group atoms removed. The corresponding calculated Omit map showed no sign of hydroxymethyl group disorder and clearly indicated that both groups were in the tg33 conformation, Figure 2. The subsequent refinement of all atomic positions, with the exception of hydrogen atoms on hydroxyl groups, with global scaling and thermal parameters, involved a total of 68 parameters, 65 restraints and resulted in values of 17.79% and 42.96% for R and Rω, respectively.34 Allowing individual thermal parameters to refine did not significantly improve the agreement with the data. The coordinates of the final model are given in the crystallographic information file supplied as Supporting Information. (30) Sheldrick, G. M., SHELX-97, a program for the refinement of Single-Crystal Diffraction Data; University of Go¨ttingen: Go¨ttingen, Germany, 1997. (31) Langan, P.; Nishiyama, Y.; Chanzy, H. Biomacromolecules 2001, 2, 410. (32) Koyama, M.; Helbert, W.; Imai, T.; Sugiyama, J.; Henrissat, B. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 9091. (33) The glycosidic torsion angles, Φ and Ψ which describe the relative orientation of adjacent glycosyl residues in the same chain are defined by (O5-C1-O1-C4) and (C1-O1-C4-C5), respectively. The glycosidic bond angle, τ is defined by (C1-O4-C4). The conformation of the hydroxymethyl group is defined by two letters, the first referring to the torsion angle χ (O5-C5-C6-O6) and the second to the torsion angle χ′ (C4-C5-C6-O6). An ideal tg conformation would be defined as the set of two angles 180°, -60°. (34) R is calculated from Σ(|Fo| - |Fc|)/Σ|Fo| with Fo > 4σ, where Fo and Fc are the observed and calculated amplitudes, respectively. Rω, is calculated from [Σω(Fo2 - Fc2)2/Σω(Fo2)2]1/2, where ω is a weight (1/σ2) applied to each F2 term in the least squares refinement. Rω will be more than twice the size of the R.
Crystal Structure and H-Bonding System in Cellulose IR
ARTICLES Table 2. Hydrogen Bonds with H‚‚‚A < R(A) + 2.800 Angstroms and 〈DHA〉 110°
Figure 3. Section through the Fourier difference synthesis calculated with coefficients (Fd - Fh)expiRc, where Fd and Fh are the observed structure factor amplitudes from deuterated and hydrogenated cellulose IR, respectively, and Rc are phases calculated from the X-ray structure reported here. The skeletal model represents the cellulose chain. The difference density is represented at two contour levels in green and magenta.
Neutron difference amplitudes, Fd - Fh, were combined with phases calculated from the X-ray structure in a Fourier difference synthesis, where Fd and Fh correspond to amplitudes collected from the deuterated and hydrogenous samples, respectively. Labile deuterium atoms on hydroxyl groups were identified with difference density peaks in the synthesis and their positions were subsequently refined using SHELX97 and previously described strategies.24 Well-defined difference density peaks, shown in Figure 3, could be clearly identified with one possible deuterium atom position on the secondary alcohol O3 atoms and two possible deuterium atom positions each for the primary alcohol O6 and secondary alcohol O2 atoms. The atom labels are as in Nishiyama et al.,24 with the exception that they are post-fixed with “u” or “d” to associate them with either one of the two independent glucosyl residues. The two partial deuterium atom positions associated with O2u, designated D2uA and D2uB, had their total occupancy constrained to unity and likewise for the partial positions associated with O2d, O6u and O6d, namely, D2dA, D2dB, D6uA, D6uB, D6dA, and D6dB, respectively. However there was no restraint imposed between deuterium atoms associated with different oxygen atoms. The resulting values for R and Rω were 20.41% and 45.33%. The refined positions of partial deuterium atoms labeled A and B corresponded to those of two intra-sheet hydrogen-bonding networks, also observed in cellulose Iβ where they were designated I and II.24 When I and II were made mutually exclusive by constraining the deuterium A atoms to have equal occupancy and the occupancy of deuterium A and B atoms to add to unity, the resulting values for R and Rω were 20.49% and 45.66% with a reduction of 3 in the number of parameters. This constrained refinement could not be rejected even at a 50% level of confidence with respect to the previous refinement, and we therefore take it as our best structure. When only A or B atoms were included in the refinement, the resulting values for R and Rω were 21.69% and 49.77%, and 23.65% and 52.53%, respectively. Both these refinements could be rejected with respect to the previous refinement at a confidence level of greater than 97.5%. The refined coordinates and occupancies of the hydrogen atoms are given in the crystallographic information file supplied as Supporting Information. The hydrogen bonding parameters are given in Table 2.
Results and Discussion
The unit cell parameters reported here are slightly smaller and, we believe, more accurate than those deduced from electron
D−H
d(D−H)
d(H‚‚A)
∠DHA
d(D‚‚A)
A
O3d-D3d O3d-D3d O2d-D2dA O2d-D2dB O2d-D2dB O2u-D2uA O2u-D2uA O2u-D2uB O2u-D2uB O6d-D6dA O6d-D6dA O6d-D6dB O6d-D6dB O6u-D6uA O6u-D6uA O6u-D6uB O3u-D3u
0.989 0.989 0.974 0.983 0.983 0.980 0.980 0.985 0.985 0.977 0.977 0.976 0.976 0.983 0.983 0.976 0.976
1.954 2.386 1.689 2.283 2.679 1.763 2.181 2.357 3.019 2.176 2.791 1.894 1.967 1.853 2.881 1.956 2.072
163.94 119.05 133.83 116.94 157.47 127.07 118.44 110.37 122.21 122.20 141.32 150.36 110.23 153.99 134.82 145.03 137.59
2.918 2.994 2.465 2.866 3.606 2.480 2.784 2.853 3.641 2.821 3.606 2.785 2.480 2.770 3.641 2.812 2.868
O5u [x-1, y, z] O1u [x-1, y, z] O6u O3d O6d [x, y-1, z+1] O6d [x+1, y, z] O1u O3u O6u [x, y+1, z-1] O3d [x, y+1, z-1] O2d [x, y+1, z-1] O1u [x-1, y, z] O2u [x-1, y, z] O3u [x, y-1, z+1] O2u [x, y-1, z+1] O1d O5d
diffraction experiments.11 The mass densities calculated from the parameters reported in this study and the electron diffraction study are 1.61 and 1.58, respectively. Electron beams are known to damage polymer crystals by inducing cross-links and chemical modifications. In the case of cellulose, it has been reported that this damage results in a swelling of the crystal cell and therefore a reduction in crystal density.35,36 Selected model parameters for IR and Iβ are given in Table 3. In Iβ, there are two conformationally distinct chains in a monoclinic P21 unit cell, referred to as the corner and center chains. Each chain lies on P21 symmetry axis that requires adjacent glucosyl residues in the same chain to be identical. In IR, there is one chain in a triclinic unit cell, and no requirement for adjacent glucosyl residues in that chain to be identical. The relative orientation of adjacent glucosyl residues can be described by the glucosidic torsion angles Φ and Ψ, the bond angle τ, and the conformation of hydroxymethyl groups can be described by torsion angles χ and χ′. 33 The values of the parameters in Table 3 are all well within the ranges observed in crystals of small analogues of cellulose.37 Whereas the conformations of the symmetry independent glucosidic linkages and hydroxymethyl groups are different in Iβ, they are identical to within experiment error in IR. However, the conformations of the symmetry independent sugar rings, represented by the calculated Cremer and Pople puckering parameter θ, are different in both IR and Iβ.38 These results are in agreement with recent CP/MAS 13C NMR studies of cellulose and cellulose derivatives. Horii et al.39,40 have shown that there is a correlation between the chemical shifts in CP-MAS spectra and the dihedral angles defined by the bonds associated with that particular carbon atom. In precisely assigning the chemical shifts of the sugar carbon atoms in IR and Iβ, it has been shown by Kono et al.41,42 that there are two nonequivalent gluco(35) Revol, J. F. J. Mater. Sci. 1985, 4, 1347. (36) Sugiyama, J.; Okano, T. In Cellulose Structure and Functional Aspects; Kennedy, J. F.; Phillips, G. O.; William P. A. Eds.; Ellis Horwood Publishers: Mancheser 1989; p. 75. (37) French, A. D.; Johnson, G. P. Cellulose, in press. (38) Cremer, D.; Pople, J. A. J. Am. Chem. Soc. 1975, 97, 1354. (39) Horii, F.; Hirai, A.; Kitamaru, R. Polym. Bull. 1983, 10, 357. (40) Horii, F.; Hirai, A.; Kitamaru, R. In Polymers for Fibers and Elastomers; Arthur Jr, J. C., Ed.; ACS Symposium Series 260, J. Am. Chem. Soc., Washington, D. C. 1984; p.27. (41) Kono, H.; Erata, T.; Takai, M. Macromolecules 2003, 36, 3589. (42) Kono, H.; Yunoki, S.; Shikano, T.; Fujiwara, M.; Erata, T.; Kawai, M. J. Am. Chem. Soc. 2002, 124, 7506. J. AM. CHEM. SOC.
9
VOL. 125, NO. 47, 2003 14303
ARTICLES
Nishiyama et al.
Table 3. Selected Model Parametersa residue 1
IR Iβ
residue 2
Φ
Ψ
τ
χ
χ′
θ
Φ
Ψ
τ
χ
χ′
θ
-98 (3) -98.5 (20)
-138 (2) -142.3 (19)
116
167 (3) 170 (3)
-75 (4) -70 (3)
9.4
-99 (3) -88.7 (20)
-140 (3) -147.1 (16)
116
166 (3) 158 (3)
-74 (4) -83 (3)
6.9
115
10.2
116
6.7
a Residue 1 corresponds to the origin chain in cellulose I and the “d” residue in cellulose I . Residue 2 corresponds to the center chain in cellulose I β R β and the “u” residue in cellulose IR. The values in parentheses are experimental standard deviations on the corresponding last digit places.
pyranosyl residues in the unit cells of both allomorphs, in agreement with the results presented here. Furthermore, Kono et al. have shown that the main differences between the IR and Iβ structures are in the conformations of the anhydroglucose residues and the β-1,4 linkages.42 The list of conformational parameters presented in Table 3 is in full agreement with the NMR results of Kono et al.42 The chemical shifts of the C6 resonance can be correlated with χ and χ′,39 and the chemical shifts associated with C1 and C4 can be correlated with Φ and Ψ, respectively.40 The fact that the C6 and C1 resonances are observed as single peaks in IR spectra but are distinctly split in Iβ spectra is in agreement with the results presented here. Because the C4 resonance is split in both IR and Iβ spectra one would expect there to be two distinct values of Ψ in ΙR and Ιβ. Although this is the case in Iβ, in IR the values of this torsion angle are equal to within experimental error. One explanation for this is that the C4 chemical shift is extremely sensitive to changes in local conformation to a degree beyond the accuracy of this study. In particular, the two symmetry independent (C1-O1-C4-C5) torsion angles -138 (3)° and -140 (3)°) could possibly differ by more than 8°. As with cellulose Iβ, there is no hint of inter-sheet O-H‚‚‚O hydrogen bonds in cellulose IR. Within each sheet the neutron Fourier difference maps indicate that although the hydrogen atoms associated with O3 occupy single well-defined positions the hydrogen atoms associated with O2 and O6 are distributed between a number of partially occupied, but still well-defined, positions. As with cellulose Iβ, these partially occupied positions can be described by two mutually exclusive hydrogen-bonding networks, designated I and II.24 It was previously noted that the situation in reality may correspond to a dynamic balance between I and II, I and II may coexist in different parts of the sample, or the situation may be more complicated with local hydrogen bonding geometries changing in both position and time. However, it is convenient to use the concept of two exclusive networks here for the purpose of comparing hydrogen bond geometries in cellulose IR and Iβ. In both cellulose IR and Iβ, there are relatively strong O3H‚‚‚O5 intrachain hydrogen bonds. In cellulose Iβ there is an alternation in these bonds between the origin and center chains (H‚‚‚A are 1.966 Å 1.752 Å, ∠DHA are 137.08°, 162.23° for origin and center chain respectively). In cellulose IR there is a similar alternation but this time down the same chain (H‚‚‚A are 2.072 Å, 1.954 Å and ∠DHA 137.59°, 163.94°). In network I, the intrachain O2-H‚‚‚O6 bond distances are shorter in IR compared to Iβ (H‚‚‚A are 1.689 Å, 1.763 Å for IR and 1.832, 1.904 in Iβ) and the interchain O6-H‚‚‚O3 bond distances are shorter in Iβ compared to IR (H‚‚‚A are 1.853 Å, 2.176 Å for IR and 1.779 Å, 2.040 Å in Iβ). However, for both of these types of bonds the angles are much closer to 180° in Iβ compared to 14304 J. AM. CHEM. SOC.
9
VOL. 125, NO. 47, 2003
Figure 4. Distribution of H‚‚‚A bond distances and ∠DHA bond angles for O2 and O6 donors in cellulose IR (blue) and Iβ (pink). Each point represents a hydrogen bond. The distance of the point from the origin represents the H‚‚‚A distance and the anticlockwise rotation of point away from the x-axis about the origin represents 180°-∠DHA
IR (∠DHA are 133.83 °, 127.07 °, 122.20 °, and 153.99 ° in IR and 158.72 °, 144.26 °, 165.12 ° and 156.61 ° in Iβ). Many of these bonds are bifurcated in IR and Iβ with similar types of minor components. In network II, there is only one instance of the intrachain O6-H‚‚‚O2 bond in IR and two in Iβ. However there are two instances of the interchain O2-H‚‚‚O6 bond in IR and only one in Iβ. The network II bond distances are shorter (H‚‚‚A are 2.679 Å, 3.019 Å and 1.967 Å in IR and 1.876 Å, 2.440 Å, and 1.967 in Iβ) and the bond angles larger (∠DHA are 157.47°, 122.21°, and 110.23° in IR and 150°, 152.06° and 135.44° in Iβ) in Iβ compared to IR. Jeffrey has used the concept of strong, moderate and weak hydrogen bonds in order to explain observed hydrogen bonding geometries. 43 The H‚‚‚A distances are 1.2-1.5 Å, 1.5-2.2 Å, and 2.2-3.2 Å for strong, medium and weak bonds and the bond angles are 175°-180°, 130°-180° and 90°-150°, respectively. Desiraju and Steiner have used two categories to describe hydrogen bonds, strong and weak, where H‚‚‚A is 1.52.2 Å and 2.2-3 Å for strong and weak bonds and the bond angles are 130°-180° and 90°-180°, respectively.44 The geometries of the major hydrogen bond components involving O2 and O6 donors in IR and Iβ are represented visually in Figure 4. Most of the hydrogen bonds are of the moderate type according to Jeffrey and the strong type according to Desiraju (43) Jeffrey, G. A. An Introduction to Hydrogen Bonding; Oxford University Press: Oxford, U.K., 1997. (44) Desiraju, G. R.; Steiner, T. The Weak Hydrogen Bond; Oxford University Press: New York, 1999.
Crystal Structure and H-Bonding System in Cellulose IR
Figure 5. Projections of the crystal structures of cellulose IR (left) and Iβ (right) down the chain axes (top), perpendicular to the chain axis and in the plane of the hydrogen bonded sheets (middle), and perpendicular to the hydrogen boned sheets (bottom). The cellulose chains are represented by red skeletal models. The asymmetric unit of each structure is also represented in thicker lines with carbons in yellow. The unit cell of each structure is shown in white.
and Steiner. However, two hydrogen bonds in IR and one hydrogen bond in Iβ are weak according to both Jeffrey and Desiraju and Steiner. It is clear from Figure 4 that the hydrogen bonds in Iβ are distributed over a region of better geometry than those in IR. The more favorable intra-sheet hydrogen bonding in Iβ must be due to the difference in cellulose chain conformation between IR and Iβ. Another difference between the hydrogen bonding in IR and Iβ is that the relative occupancies of the two networks are different (network I is ∼70-80% occupied in Iβ and ∼55% in IR). The significance of this was tested by repeating the cellulose IR deuterium atom refinement, this time constraining the deuterium A atoms to have an occupancy of 75% as in cellulose Iβ. The resulting values for R and Rω were 20.79% and 46.63%, respectively. This refinement could be rejected with respect to the previous refinements at a confidence level of greater than 97.5%. In summary, the intra-sheet hydrogen bonding involving O2 and O6 donors is disordered over two possible networks in both cellulose IR and Iβ. The geometry of these networks is better in Iβ than in IR. The occupancy of these networks is different in IR and Iβ. The strong intrachain O3-H‚‚‚O5 hydrogen bonds that give the cellulose chains their ribbon shape alternate between two slightly different geometries in both cellulose IR and Iβ. In
ARTICLES
cellulose IR the alternation is down the same chain. In cellulose Iβ the alternation is between conformationally distinct chains. The projections of the crystal structures of cellulose IR and Iβ down the chain axes are remarkably similar, Figure 5. As the projection perpendicular to the chain axis and in the plane of the hydrogen bonded sheets shows, the main difference between IR and Iβ is the relative displacement of the sheets in the chain direction. In both IR and Iβ, the second sheet, designated II, is shifted in the “up” direction by ∼c/4 relative to the first sheet, designated I. The third sheet, designated III, is similarly shifted with respect to II by ∼c/4 in IR but in Iβ it is shifted by a ∼c/4 in the “down” direction. There is a relative difference of ∼c/2 in the position of III with respect to II in IR and Iβ. Because there exists an approximate molecular 21 screw axis, this difference is equivalent to stacking opposite faces of III on II. The most likely route for solid-state conversion of cellulose IR f Iβ is the relative slippage of the cellulose chains past one another.45 This movement does not require the disruption of the hydrogen-bonded sheets (along the 100 planes for cellulose Iβ and 110 planes for cellulose IR) but slippage by ∼c/2 at the interface of sheets of type II and III in Figure 5. We note that there is also the possibility that the chains in III could rotate around their axes, because a slippage of ∼c/2 is equivalent to a rotation of 180° around a 21 molecular axis. However this seems unlikely. It has been hypothesized that the inner cohesion of ambient cellulose crystals is largely due to the van der Waals attraction between hydrogen-bonded sheets.46,47 Despite the apparent strength of these interactions at room temperature, they are clearly effected by heating the crystals; the lattice is remarkably constant in directions within the hydrogen-bonded sheets, but expands in the stacking direction.48 Wada has recently observed a new phase in cellulose at temperatures higher than 220/230 °C, where the conversion IR f Iβ takes place.49 Little is known about this phase except that it corresponds to a crystalline structure where the inter-sheet distance has expanded by about 6%, whereas distances within the sheets are essentially constant. A recent study of R-glycine suggests a possible mechanism for this anisotropic thermal expansion. In crystals of R-glycine, molecules are arranged in hydrogen-bonded sheets with the molecular axes lying within the sheets.50 On heating, the unit cell expands anisotropically with the stacking direction increasing the most. The increasing separation of the sheets is driven by an increase in molecular libration about an axis that lies in the plane of the sheets and which causes the movement of atoms out of this plane. There is a corresponding weakening of the inter-sheet bonding interactions. It is possible that librations of the cellulose chains about axes that lie in the plane of the hydrogen-bonded sheets is the driving mechanism for sheet separation, and lubrication, during conversion of IR f Iβ. Although van der Waals attraction between hydrogen-bonded sheets may well contribute to cellulose crystal cohesion our results indicate that weak C-H‚‚‚O hydrogen bonding also plays (45) Hardy, B. J.; Sarko, A. Polymer 1996, 37, 1833. (46) French, A. D.; Miller D. P.; Aabloo, A. Int. J. Biol. Macromol. 1993, 15, 30. (47) Cousins, S. K.; Brown, R. M. Jr. Polymer 1995, 36, 3885. (48) Wada, M.; Kondo, T.; Itoh, T. Polym J. 2003, 35, 155. (49) Wada, M. J. Polym. Sci., Part B: Polym. Phys. 2002, 40, 1095. (50) Langan, P.; Mason, S. A.; Myles, D.; Schoenborn, B. P. Acta Crystallogr. 2002, B58, 728. J. AM. CHEM. SOC.
9
VOL. 125, NO. 47, 2003 14305
ARTICLES
Nishiyama et al.
Table 4. Potential Inter-Sheet Weak Hydrogen Bonds with D‚‚‚A < r(D) + r(A) + 0.50 Å D(H‚‚‚A) < r(H) + r(A)-0.12 Å and D-H‚‚‚A > 100.0°a
IR IR IR IR Iβ Iβ Iβ Iβ Iβ
D
H
A
H‚‚‚A
D‚‚‚A
D−H‚‚‚A (deg)
C2u C6d C6d C6u C1c C2c C3o C5c C6c
H2u H6Au H6Bu H6Au H1c H2c H3o H5c H6Ac
O1d O2d O2d O5d O6o O3o O2c O1o O2o
2.5136 2.5078 2.5842 2.4738 2.5476 2.4930 2.5158 2.5862 2.3585
3.40(4) 3.48(5) 3.43(5) 3.23(4) 3.3908 3.4668 3.4612 3.4353 3.3232
149.86 176.31 146.36 134.63 144.15 172.39 162.01 145.01 172.84
a The values in parentheses are experimental standard deviations on the corresponding last digit places.
a role. Table 4 shows that there are potential weak hydrogen bonds of this type in both IR and Iβ. There are more C-H‚‚‚O inter-sheet bonds in Iβ than in IR. The average H‚‚‚A distances are about the same in IR (2.52 Å) and Iβ (2.50 Å). The bond angles for such weak interactions are probably not so important but they are larger in Iβ (159.28°) than in IR (151.79°). It is likely that enhanced C-H‚‚‚O interactions contribute to the stability of Iβ over IR. Both Iβ and IR also contained short H‚‚‚H interactions (2.52 Å in Iβ and 2.4 Å and 2.46 Å in IR). No attempt was made to relieve these contacts because they are typical for short H‚‚‚H distances involved in single-crystal studies of cellulose fragments. As discussed above, the crystallographic structures and hydrogen bonding arrangements for cellulose IR and cellulose Iβ provide important insights into cellulose stability and transformation and contribute toward a scientific basis for understanding cellulose biogeneration and reactivity. However, they also provide the basic parameters required for the application of a number of other techniques to understanding the properties of cellulose. In particular, these results will enable modeling studies of the properties of whole fibers of cellulose, IR/Iβ interfaces within those fibers and the surface properties of
14306 J. AM. CHEM. SOC.
9
VOL. 125, NO. 47, 2003
those fibers. These results also provide an important key for interpreting spectroscopic data. In addition to their practical impact, these results have brought some surprises. In the IR phase, the cellulose molecules adopt a conformation remarkably close to a 2-fold screw, which is not required by the triclinic space group. On the other hand, in the Iβ allomorph, where the 2-fold screw is required by the symmetry, surprising conformational differences exist between the center and the corner chains. The presence of two similar hydrogen bonding networks with different occupancies in cellulose IR and Iβ was not expected. Neither was it expected that there would be an alternation of intrachain O3-H‚‚‚O5 hydrogen bonds in both phases. At the same time these crystallographic studies were underway, 2D solid-state NMR studies were being carried out by Kono et al on cellulose IR and Iβ.42Although these studies were done independently and simultaneously, they have provided remarkably consistent information about cellulose conformation at an unprecedented level of detail. At the start of these studies, not many would have projected that cellulose IR and cellulose Iβ differ to such an extent in crystal packing, molecular conformation and hydrogen bonding. Acknowledgment. The authors thank the Institut Laue Langevin and the European Synchrotron Radiation Facilities for the provision of beamtime. Y.N. thanks the French Government and the Japanese Society for the Promotion of Science for financial support. P.L. thanks the Office of Science and the Office of Biological and Environmental Research of the US Department of Energy for financial support. Supporting Information Available: Coordinates of the final model and the refined coordinates and occupancies of the hydrogen atoms are given in a crystallographic information file. This material is available free of charge via the Internet at http://pubs.acs.org. JA037055W